NVIDIA Commits $100 Billion to OpenAI in Landmark AI Infrastructure Deal

NVIDIA has committed $100 billion to OpenAI in a landmark agreement that will channel billions into the startup’s fund while primarily financing advanced NVIDIA processors to power future AI data centers. The investment will be disbursed in phases as supercomputing facilities go online, with the first slated for 2026. Instead of buying GPUs outright, OpenAI plans to lease them, a structure that lowers upfront costs but shifts more risk to NVIDIA. The arrangement begins with an initial $10 billion transfer to OpenAI, intended to support the development of its first gigawatt-scale facility.

NVIDIA CEO Jensen Huang estimated that a single gigawatt AI data center could cost about $50 billion, of which roughly $35 billion would be allocated to GPUs. While a portion of the investment may be used for hiring and operations, the majority is expected to fund computing power central to training large AI models. For OpenAI, which lacks both investment-grade credit and positive cash flow, this financing model offers a more affordable route to securing chips while also improving its prospects of obtaining bank loans for additional expansion.

The deal underscores NVIDIA’s rise, with its valuation climbing to $4.3 trillion on surging GPU demand from OpenAI, Microsoft, Meta, Google, and Amazon. OpenAI, now valued near $500 billion, has relied heavily on external capital, much of which has been reinvested in the very companies supplying its infrastructure. Analysts caution that this cycle of raising and reinvesting inflates figures without addressing underlying financial risks. OpenAI CEO Sam Altman countered that the company remains focused on building products and services that generate sustained customer demand, which in turn will continue to require significant data center and chip investments.